Robert M.
La Follette School of Public Affairs
at the University of Wisconsin-Madison
Working Paper Series
La Follette School Working Paper No. 2015-006
http://www.lafollette.wisc.edu/research-public-service/publications
(Almost) A Quarter Century
of Currency Expectations Data:
Interest Rate Parity and the Risk Premium
Menzie D. Chinn
Professor, La Follette School of Public Affairs and Department of Economics
at the University of Wisconsin-Madison, and National Bureau of Economic Research
mchinn@lafollette.wisc.edu
April 2015
1225 Observatory Drive, Madison, Wisconsin 53706
608-262-3581 / www.lafollette.wisc.edu
The La Follette School takes no stand on policy issues; opinions expressed
in this paper reflect the views of individual researchers and authors.
(Almost) A Quarter Century of Currency Expectations Data:
Interest Rate Parity and the Risk Premium
by
Menzie D. Chinn
University of Wisconsin, Madison
and NBER
September 28, 2014
Abstract: I re-examine whether the presence of an exchange risk premium or biased
expectations explains why the forward premium fails to predict exchange rates changes.
The re-examination is undertaken using survey-based data on exchange rate expectations
over a long sample extending from August 1986 to October 2009 period. I find that on a
time series basis, (i) forward rate bias persists into the most recent period, (ii) the forward
rate better predicts expected depreciation, suggesting uncovered interest parity holds, (iii)
these patterns persist in panel regressions, and (iv) survey based expectations are biased
predictors of exchange rate changes. The implication is that the standard measure of the
exchange risk premium, identified using the rational expectations hypothesis, provides
misleading inferences.
Keywords: forward rate unbiasedness, efficient markets hypothesis, risk premium,
survey data, rational expectations.
Acknowledgements: I thank participants at the 2012 ASSA meetings and seminar
participants at the University of Houston for useful comments. Yi Zhang, Saad Quayyum,
Jonathan McBride and Kieran Coe provided research assistance. I acknowledge the
support of faculty research funds of the University of Wisconsin and from the UW Center
for World Affairs and the Global Economy for financial support.
───────────────────────────────────────────────────────────
Correspondence: Robert M. LaFollette School of Public Affairs; and Department of
Economics, University of Wisconsin, 1180 Observatory Drive, Madison, WI
53706-1393. Email: mchinn@lafollette.wisc.edu .
1. INTRODUCTION
One of the key puzzles in international macroeconomics is the failure of the forward
rate to predict future movements in the spot exchange rate. Another way of putting this is that, if
covered interest parity holds, then interest differentials fail to predict the direction of exchange
rate changes. This seeming violation of uncovered interest parity is one of the most robust
stylized facts in the discipline. While this outcome can be explained by appealing to the presence
of an exchange risk premium, the difficulty in relating measured risk premiums to observable
macroeconomic phenomena has meant that dispensing with one puzzle leads to yet another
puzzle.
I emphasize the word “seeming” because in fact most empirical papers assessing
uncovered interest parity are actually joint tests of uncovered interest parity and the rational
expectations hypothesis. Frankel has termed this composite the “unbiasedness hypothesis”.
The
forward rate unbiasedness hypothesis is consistent with the UIP, rational expectations and
covered interest parity (so that the forward discount equals the interest differential). These
distinctions, while straightforward, are critical for understanding why the forward rate might not
predict the future spot rate. It could be because of an exchange risk premium; or it could be
because expectations are not on average unbiased.
This paper, I eschew the approach of assuming the rational expectations hypothesis,
and instead use survey based measures of exchange rate expectations to proxy for market
expectations. Early contributions in this vein were Dominguez (1986), Frankel and Froot (1987),
1
and Froot and Frankel (1989).1 The empirical results presented in this paper are based on a data
set derived from FXForecasts, the successor to Currency Forecasters' Digest and Financial
Times Currency Forecaster. This data set has the advantage of spanning nearly a quarter of a
century (23 years) for several of the currencies. To my knowledge, this is the longest sample
period over which survey data has been used to analyze the foreign exchange market.
To anticipate the results, I find that the forward discount does positively correlate with
expected depreciation, in a manner consistent with uncovered interest parity. In contrast, the
usual relationship holds for ex post exchange rate changes, over the corresponding sample
periods – that is interest differentials tend to point in the wrong direction for future changes in
exchange rates. Panel regression results from the two approaches are less divergent, but even in
these regressions, interest differentials predict incorrectly the direction of change in exchange
rates at the year horizon.
These results are consistent with biased exchange rate expectations. I show that for
many cases (particularly where the results differ substantially between regressions using the
actual changes and expected changes) the exchange rate bias is significant.
I also find that the exchange risk premium identified using survey data behaves much
differently than that implied by rational expectations. This is a finding that is more clearly
highlighted when using a longer sample period. In particular, the risk premia based on survey
data are much more persistent than the risk premia obtained using the conventional approach.
Furthermore, the evidence suggests negative risk premia for the Japanese yen and Swiss franc
(relative to the US dollar).
1
Takagi (1991) reviewed the early literature. Chinn and Frankel (1994), and Frankel and Chinn (1993) use this data
source.
2
This paper is organized in the following fashion. In section 2, I discuss the uncovered
interest parity condition, combined with the rational expectations hypothesis (sometimes called
the risk neutral efficient markets hypothesis, or “RNEMH”), and in section 3, UIP is evaluated
empirically, under the conventional rational expectations hypothesis as well as the case where
survey data are used to measure expectations. Section 4 examines why these differing results
might arise. Section 5 provides some panel estimates, while Section 6 concludes.
2. Uncovered Interest Parity, the Unbiasedness Hypothesis and the Risk Neutral Efficient
Markets Hypothesis
Let st be the price of foreign currency in units of domestic currency at time t, ft,t+k is the
forward value of s for a contract expiring k periods in the future (both in logs). Suppose the
forward rate (in logs, f) differs from the expected future spot rate (denoted by the e superscript)
by a premium that compensates for the perceived riskiness of holding domestic versus foreign
assets. The risk premium, η, is defined as:
f t,t+k = s et,t+k +  t+k .
(1)
Subtracting the log spot rate at time t from both sides, and rearranging yields:
s
e
t,t +k
 s t   f t,t+k  s t    t+k .
(2)
Expected depreciation equals the forward discount, minus the risk premium. Note that if covered
interest parity holds,
(3)
f t,t+k - st = ( it,k - i*t,k ) .
and the risk premium is zero, then equation (2) becomes the familiar uncovered interest parity
condition:
3
 s t,et+k = ( i t,k - i*t,k )
(4)
Where it+,k is the k-period yield on the domestic instrument, and i*t+k is the corresponding yield
on the foreign instrument.
The forward discount equals expected depreciation if the risk premium is zero.2 This is
sometimes termed the forward rate efficient markets hypothesis. Equations (2) and (4) are not
directly testable, however, in the absence of observations on market expectations of future
exchange rate movements. To make this hypothesis testable, it is tested jointly with the
assumption of rational expectations. Using the rational expectations methodology, future
realizations of st+k will equal the value expected at time t plus a white-noise error term ξt+k that is
uncorrelated with all information known at t, including the interest differential and the spot
exchange rate:
st+k = s ret,t+k +  t+k ,
(5)
Where the “re” superscript denotes the rational expectations measure. Then, applying the
expression (2) one obtains the following relationship,
 st , t+k =  f t,t+k  s t  -  t,t+k +  t+k ,
(6)
where the left-hand side of equation (6) is the realized change in the exchange rate from t to t+k.
According to the forward rate efficient markets hypothesis, the risk premium is zero and both the
risk premium and the error term are assumed to be orthogonal to the interest differential.
In a regression context, the estimated parameter on the forward premium will have a
probability limit of unity in the following regression:
2
Note that some approximations and simplifying assumptions have been made in order to arrive at this expression.
See Engel (1996).
4
 s t , t +k =  + 
f
t,t +k
 s t  +  t +k .
(7)
If the joint hypothesis holds, then the disturbance in equation (7) becomes simply the rational
expectations forecast error ξt,t+k, which by definition is orthogonal to all information known at
time t, including the forward discount.
Forward rate unbiasedness is a weaker condition than the risk neutral efficient markets
hypothesis. All that is required for forward rate unbiasedness is that any risk premium and/or
non-rational expectations error be uncorrelated with the forward discount, while the risk neutral
efficient markets hypothesis requires in addition that no other regressors known at time t should
have explanatory power.3
Estimates of equation (7), assuming covered interest parity, using values for k that range up
to one year typically reject the unbiasedness restriction on the slope parameter. For instance, the
survey by Froot and Thaler (1990), for instance, finds an average estimate for β of -0.88.4
One can relax the assumption regarding expectations, and replace it with another: that
survey based expectations are a good measure for market expectations. Hence, instead of
equation (7), estimate.
sˆte,t  k =  ' +  '  f t,t+k  st + ~t+k .
(8)
where ste,t  k  ste,t  k  st is the expected depreciation implied by survey data. Under the null
hypothesis of uncovered interest parity, the probability limit of β’ equals unity as long as the as
long as error term is uncorrelated with the interest differential.
Froot and Frankel (1989) demonstrate that the standard tests for forward rate bias yield
3
The constant term may reflect a constant risk premium demanded by investors on foreign versus domestic assets.
Default risk could play a similar role, although the latter possibility is less familiar because tests of UIP (as well as
CIP) generally use returns on assets issued in offshore markets by borrowers with comparable credit ratings.
4
Similar results are cited in surveys by MacDonald and Taylor (1992) and Isard (1995).
5
radically different results when one uses survey-based measures of exchange rate depreciation.
They find that most of the variation of the forward discount appears to be related to expected
depreciation, rather than a time varying risk premium, thereby lending credence to UIP. [Since
covered interest parity holds for these currencies, the forward discount is equivalent to the
interest differential].
Chinn and Frankel (1994) document the fact that it is difficult to reject the forward
rate unbiasedness hypothesis for a broader set of currencies, when using forecasts provided by
the Currency Forecasters’ Digest (CFD), although there is some evidence of a risk premium at
the 12 month horizon. Chinn and Frankel interpret the differing results as arising from a wider
set of currencies – they examine 17 currencies as opposed to the 5 or so examined by Frankel
and Froot – where the assumption of perfect substitutability of debt instruments is less likely to
hold.
Note that the object I identify as the risk premium need not be a true exchange risk
premium, such as that identified by Bacchetta and van Wincoop (2009). In that case, infrequent
portfolio decisions account for the gap between the forward rate and the expected spot rate.
3. Empirics
In this section, I compare the results from the standard unbiasedness tests and the test for
UIP using survey data.
3.1 Unbiasedness
First, consider the results of estimating equation (7):
 s t , t +k =  + 
f
t,t +k
 s t +  t +k .
(7)
6
Table 1 reports the results from estimating the standard ex post UIP regression (UIP
incorporating rational expectations), often known as the “Fama regression” (1984). While data
are available at the 1, 3, 6 and 12 month horizons, only results for the three and 12 months
horizons are reported. Under the maintained hypothesis, the errors should be serially
uncorrelated at the one month horizon. At the multi-month horizons, even under the null of
rational expectations, there should be moving average serial correlation of order k-1 (Hansen and
Hodrik, 1980), i.e., order 2 and order 11 for the three month and 12 month horizon regressions,
respectively. However, we report the estimates using Newey-West standard errors, as there
appears to be serial correlation, according the Durbin Watson statistics, above and beyond that
implied by overlapping horizons.
In the top panel of Table 1.1 and 1.2 are presented the estimates for the euro legacy
currencies (as well as the euro). For the legacy currencies, the sample ends in such a way that the
last forecasted exchange rate is 1998M12. That means that at the three month horizon, the
sample ends at 1998M09. For the euro, the sample begins at 1999M01 and ends at 2009M10
(and thus incorporates forecasts of the euro in 2010M01). The results are much in line with those
reported elsewhere in the literature. Slightly over half the point estimates are negative; despite
this, the standard errors are so large that one can reject the null of a coefficient of unity less than
half of the time.
On the other hand, the positive coefficients are associated with the currencies of Ireland,
Italy and Spain – countries that exhibited relatively high inflation during the sample period. This
finding is consistent with Chinn and Meredith’s (2004) conclusion that the currencies of the
higher inflation countries tended to conform to the unbiasedness hypothesis at short horizons. In
7
this earlier sample, all the adjusted R-squared statistics are quite low.
The bias is also evident for the newest currency in the data set – the euro. In this case, the
imprecision of the estimates is sufficiently large at the 3 month horizon that one cannot reject the
null of a coefficient of unity at any horizon. However, the point estimate is very negative,
sufficiently so that one can reject unbiasedness at the 12 month horizon.
Panel 2 of Tables 1.1 and 1.2 reports the estimates for currencies estimated over the full
sample. The Swedish krone at the three month horizon is the lone instance where the coefficient
is above unity. Otherwise, the slope coefficients are below one, and particularly at the 12 month
horizon, significantly so.
An interesting result is that the point estimates are quantitatively close to the posited value of
unity in two cases – Sweden and Spain. Italy’s coefficients at the short horizon is very high, in
excess of 2. It is interesting in the latter two countries’ currencies, the rate of the inflation over
the sample period (which ends in 1998M12) is the highest. This result is consistent with the
findings in Chinn and Meredith (2004). High inflation currencies tend to exhibit positive
coefficients in unbiasedness regressions, equivalent here to the Fama regression.
3.2 Uncovered Interest Parity
The results of estimating ex ante uncovered interest parity stand in stark contrast to those
from ex post UIP. In order to implement the tests, we use extended versions of the data used in
Chinn and Frankel (1994). These are data provided by FX Forecasts, the successor organization
to Currency Forecaster’s Digest, and the data used are at the 3 and 12 month horizons.
Table 2 presents the results. The most obvious and striking difference is that there is only
one negative estimated coefficient for all the currencies (Japan at the 3 month horizon). In all
8
other instances, the estimated coefficients are positive, and in most cases reject the null of zero.
On the other hand, one can reject the null hypothesis of a unit coefficient consistent with the
UIP hypothesis in only 8 cases (of which four cases pertain to the situation where the point
estimate is above unity).
If one wants to focus on the major currencies, such as the Deutschemark and the euro, the
UK pound, Swiss franc, and Japanese yen, one finds that in almost all instances, one can’t reject
the null of a unit coefficient, with the exception of the Japanese yen and UK pound. So for key
currencies, UIP does seem to hold.
Why do the results differ to so widely between each approach to expectations? One can
examine this from a mechanical perspective. If exchange rate expectations, as measured by the
survey data, point in a substantially different direction from the actual exchange rate changes,
then one would expect differing results. One can quantify the differences by examining whether
expected changes exhibit bias.
 st , t+k =  +  ( sˆt ,t  k ) + u t+k .
(9)
These results are reported in Tables 3.1 and 3.2, for the 3 month and 12 month horizons,
respectively. One interesting fact is that almost all the survey-based forecasts are biased, and
exhibit very small correlation with the actual exchange rate changes. However, it is also notable
that in most of the cases where the beta coefficients switch from negative to positive are the
instances where the survey-based expected changes are negatively correlated with the actual
changes.
Another point of commonality with the rational expectations-UIP hypothesis is that the
proportion of variation explained is very low, with the exception of the 12 month horizon.
9
Morever, a high degree of serial correlation is evident in both the unbiasedness and and UIP
regressions.
3.3 Panel Regressions
The basic outlines of these results persist when undertaking panel analysis. In general, the slope
coefficients obtained using the rational expectations hypothesis are less than those obtained using
survey data. For the euro legacy currencies at the three month horizon, the contrast is the
strongest. The unbiasedness regression coefficient estimates are negative, while the UIP
regression coefficients are positive; there is a difference also for the twelve month horizon.
The contrast disappears when examining non-euro currencies over the full sample at the
three month horizon. Yet, it appears at the twelve month horizon. There the difference is
statistically and numerically different.
Overall, the panel regressions indicate that there is substantial evidence against forward
rate unbiasedness, using either the rational expectations assumption, or using survey data as
proxies for market expectations. At the twelve month horizon, the evidence is more in accord
with forward rate unbiasedness.
4. The Risk Premium
In simple models, the exchange risk premium arises from the correlation of currency returns with
the marginal utility of consumption. Of course, numerous researchers have failed to relate the
risk premium identified using rational expectations to macroeconomic fundamentals.5
Here, I examine how the risk premium defined under rational expectations differs from
5
See Engel (1996) for a discussion, and Engel (2011) for a recounting of what attributes the risk premium must
fulfill.
10
that defined using survey data. The three month risk premia are illustrated in Figure 1. The blue
line presents the risk premia obtained using survey data, while the red line depicts the
conventional risk premia implied by the rational expectations hypothesis. Clearly, the risk premia
obtained using the survey data are much more persistent than the rational expectations implied;
they also exhibit much less high frequency volatility.
To formally quantify the degree of persistence, I sampled the three month risk premia every
three months (end of each quarter), so as to eliminate the overlapping data issue. The results of
regressing these sampled data on the lagged risk premia are presented in Table 5.
The pattern is striking. In almost every case, the risk premium obtained using survey data is
highly persistent, while that obtained using the rational expectations is not. In fact it would be
hard to distinguish the latter from white noise; the AR(1) coefficient for the rational expectations
derived risk premium is essentially zero. In contrast, the half-life of a typical survey-based risk
premium is about 2 quarters.
It’s conceivable that the earlier stylized fact that the exchange risk premium is unrelated to
macroeconomic variables is in fact an artifact of the questionable assumption of rational
expectations. I reserve that question for future research.
5. Conclusions
In this study, I have re-examined the forward rate uncovered interest parity hypothesis,
using the rational expectations hypothesis, and using survey data, to identify expected exchange
rate changes. I arrive at the following conclusions:

Forward rate unbiasedness is generally generally rejected a currency by currency basis
11
using the rational expectations hypothesis.

The forward discount deviates from expected depreciation in about a third of the
currencies when using survey data based expectations. Assuming covered interest parity
holds, this means the interest differential does on average predict correctly the direction
of expected exchange rate changes. Nonetheless, one can still reject the null of uncovered
interest parity in many cases, particularly at the three month horizon.

Oftentimes, the difference in the results is linked to the finding of bias in exchange rate
expectations. This finding suggests that biased expectations are an important reason why
the forward discount (and hence the interest differential) point in the wrong direction for
subsequent ex post exchange rate changes.

Panel regression estimates indicate a failure to reject the risk neutral efficient markets
hypothesis at the three month horizon, while the slope coefficient is always positive using
survey data.

The risk premium identified using survey data differs substantially in terms of persistence
and high frequency volatility from the standard risk premium. The survey-data based risk
premium is much more persistent.
Future work should test for heterogeneity across major vs. minor currencies, and
whether pooling across currencies is justified. In addition, I plan to investigate whether the risk
premium identified using survey data is linked to any macro variables, including those suggested
by theory (consumption growth, inflation, etc.).
12
REFERENCES
Bacchetta, Philippe, and Eric van Wincoop. 2010. “Infrequent Portfolio Decisions: A Solution to
the Forward Discount Puzzle,” American Economic Review 100(3): 870-904.
Bacchetta, Philippe, Elmar Mertens, and Eric van Wincoop. 2009. “Predictability in financial
markets: What do survey expectations tell us?” Journal of International Money and Finance 28:
406–426.
Chinn, Menzie and Jeffrey Frankel. 1994. "Patterns in Exchange Rates for 25 Currencies."
Journal of Money, Credit and Banking 26(4): 759-770.
Dominguez, Kathryn. 1986."Are Foreign Exchange Forecasts Rational? New Evidence from
Survey Data?" Economics Letters 21: 277-82.
Engel, Charles, 1996, “The Forward Discount Anomaly and the Risk Premium: A Survey of
Recent Evidence,” Journal of Empirical Finance. 3 (June): 123-92.
Engel, Charles, 2011, “The Real Exchange Rate, Real Interest Rates, and the Risk Premium,”
mimeo, May 31, 2011.
Fama, Eugene. 1984."Forward and Spot Exchange Rates." Journal of Monetary Economics 14:
319-38.
Frankel, Jeffrey and Menzie Chinn. 1993."Exchange Rate Expectations and the Risk Premium:
Tests for a Cross Section of 17 Currencies." Review of International Economics1(2): 136-144.
Frankel, Jeffrey and Kenneth Froot. 1987."Using Survey Data to Test Standard Propositions
Regarding Exchange Rate Expectations." American Economic Review 77(1) (March): 133-153.
Reprinted in Frankel, On Exchange Rates, MIT Press, 1993.
Froot, Kenneth and Jeffrey Frankel. 1989."Forward Discount Bias: Is It an Exchange Risk
Premium?" Quarterly Journal of Economics. 104(1) (February): 139-161.
Froot, Kenneth and Richard Thaler. 1990."Anomalies: Foreign Exchange." Journal of Economic
Perspectives 4(3) (Summer): 179-192.
Goodhart, Charles. 1988."The Foreign Exchange Market: A Random Walk with a Dragging
Anchor." Economica 55(220) (November): 437-460.
Hansen, Lars and Robert Hodrick. 1980."Forward Exchange Rates As Optimal Predictors of
Future Spot Rates: An Econometric Analysis." Journal of Political Economy 88(5): 829-853.
13
Hodrick, Robert J. 1987. The Empirical Evidence on the Efficiency of Forward and Futures
Foreign Exchange Markets. (Chur: Harwood Academic Publishers).
Hodrick, Robert J. and Sanjay Srivastava. 1986. "An Investigation of Risk and Return in
Forward Foreign Exchange." Journal of International Money and Finance 3: 5-30.
Ito, Takatoshi. 1990. "Foreign Exchange Rate Expectations: Micro Survey Data." American
Economic Review 80 (June): 434-49.
Lewis, Karen. 1989."Changing Beliefs and Systematic Rational Forecast Errors with Evidence
from Foreign Exchange." American Economic Review 79(4) (Sept.): 621-636.
Takagi, Shinji. 1991. "Exchange Rate Expectations: A Survey of Survey Studies." IMF Staff
Papers 38(1) (March): 156-183.
14
Table 1.1. Unbiasedness Regressions, Three Month Horizon
coefficient constant beta Obs. adj.R Sq. DW coefficient constant beta Obs. adj.R Sq. DW BE FR GY IR
IT
NE
SP
EU 0.011 0.032 ‐0.551 0.912 0.008 0.030 ‐0.010 0.854 0.008 0.031 ‐0.243 0.283 0.039
0.031
0.115
1.063
‐0.055
0.061
2.477
1.819
0.008
0.031
‐0.115
0.999
‐0.037
0.053
1.509
1.284
0.003 0.034 0.220 1.315 146 ‐0.002 0.542 146 ‐0.007 0.546 146 ‐0.003 0.568 146
‐0.007
0.573
146
0.067
0.665
146
‐0.007
0.562
146
0.032
0.706
130 ‐0.007 0.677 DE NO SN SW
UK
JP
AU
CA NZ
0.008 0.023 ‐0.051 0.705 0.013 0.027 0.366 0.291 0.016 0.029 1.562 0.453 0.001
0.029
‐0.403
0.855
0.002
0.022
0.956
0.659
‐0.065
0.025
‐2.358
0.855
0.034
0.029
‐0.378
0.965
0.004 0.019 ‐0.492 0.565 0.023
0.045
‐0.353
1.508
279 ‐0.004 0.614 279 0.001 0.615 279 0.140 0.607 279
‐0.002
0.604
279
0.010
0.647
279
0.049
0.597
279
‐0.003
0.834
279 ‐0.001 0.771 250
‐0.003
0.556
Notes: OLS regression estimates; Newey-West standard errors. Top panel, 1986M08-1998M09, except for EU, 1999M01-2009M10.
Bottom panel, 1986M08-2009M10, except for NZ, 1989M01-2009M10. Entries in bold face denote significance at the 5% level, for
null hypothesis of beta=1.
15
Table 1.2. Unbiasedness Regressions, Twelve Month Horizon
coefficient constant beta Obs. adj.R Sq. DW coefficient constant beta Obs. adj.R Sq. DW BE FR GY IR
IT
NE
SP
EU 0.002 0.017 ‐0.965 0.808 0.002 0.018 ‐0.148 0.719 ‐0.002 0.017 ‐0.323 0.602 0.006
0.018
0.102
0.741
‐0.016
0.025
1.283
0.778
‐0.002
0.017
‐0.549
0.642
‐0.013
0.029
0.785
0.768
‐0.021 0.019 ‐1.963 1.265 146 0.034 0.186 146 ‐0.006 0.193 146 0.000 0.197 146
‐0.006
0.176
146
0.086
0.244
146
0.012
0.187
146
0.032
0.157
130 0.064 0.265 DE NO SN SW
UK
JP
AU
CA NZ
‐0.001 0.014 ‐0.527 0.637 ‐0.001 0.015 0.054 0.467 0.003 0.017 0.697 0.644 ‐0.034
0.019
‐1.301
0.700
‐0.003
0.016
0.404
0.930
‐0.091
0.014
‐2.492
0.534
0.018
0.022
‐0.875
0.545
‐0.008 0.011 ‐0.017 0.552 ‐0.000
0.039
‐0.664
1.425
279 0.010 0.190 279 ‐0.003 0.185 279 0.022 0.150 279
0.059
0.227
279
0.001
0.181
279
0.214
0.284
279
0.026
0.185
279 ‐0.004 0.181 250
0.001
0.124
Notes: OLS regression estimates; Newey-West standard errors. Top panel, 1986M08-1997M12, except for EU, 1999M01-2009M10.
Bottom panel, 1986M08-2009M10, except for NZ, 1989M01-2009M10. Entries in bold face denote significance at the 5% level, for
null hypothesis of beta=1.
16
Table 2.1. Uncovered Interest Parity Regressions, Three Month Horizon
coefficient constant beta Obs. adj.R Sq. DW coefficient constant beta Obs. adj.R Sq. DW BE FR GY IR
IT
NE
SP
EU 0.032 0.018 0.956 0.439 0.036 0.018 0.825 0.477 0.033 0.017 0.609 0.258 ‐0.016
0.018
0.990
0.505
‐0.002
0.021
1.309
0.346
0.036
0.017
1.659
0.431
0.032
0.027
0.326
0.452
‐0.019 0.010 0.628 0.267 141 0.071 0.366 140 0.048 0.374 141 0.112 0.379 141
0.079
0.455
141
0.141
0.433
141
0.166
0.309
141
0.004
0.410
128 0.007 2.255 DE NO SN SW
UK
JP
AU
CA NZ
0.002 0.011 0.953 0.342 ‐0.027 0.013 0.857 0.267 0.001 0.011 0.198 0.150 0.034
0.012
1.243
0.344
0.018
0.011
0.176
0.226
0.010
0.015
‐0.473
0.415
‐0.053
0.010
1.781
0.189
‐0.012 0.005 0.591 0.188 0.245
0.087
3.801
2.324
272 0.055 1.272 235 0.153 0.857 272 0.028 0.666 272
0.071
1.263
272
0.000
0.959
272
0.007
0.654
272
0.188
1.430
272 0.018 2.282 243
0.016
0.439
Notes: OLS regression estimates; Newey-West standard errors. Top panel, 1986M08-1998M09, except for EU, 1999M01-2009M10.
Bottom panel, 1986M08-2009M10, except for NZ, 1989M01-2009M10. Entries in bold face denote significance at the 5% level, for
null hypothesis of beta=1.
17
Table 2.2. Uncovered Interest Parity Regressions, Twelve Month Horizon
coefficient constant beta Obs. adj.R Sq. DW coefficient constant beta Obs. adj.R Sq. DW BE FR GY IR
IT
NE
SP
EU 0.042 0.007 1.078 0.214 0.038 0.007 0.960 0.224 0.045 0.006 1.412 0.204 ‐0.029
0.008
0.803
0.293
0.012
0.009
1.129
0.179
0.043
0.007
1.294
0.202
0.019
0.010
0.657
0.199
‐0.029 0.008 1.333 0.358 139 0.223 0.468 140 0.183 0.434 141 0.419 0.486 141
0.163
0.542
139
0.288
0.487
141
0.332
0.401
139
0.154
0.461
128 0.142 0.838 DE NO SN SW
UK
JP
AU
CA NZ
‐0.002 0.007 1.417 0.231 ‐0.018 0.007 0.842 0.243 ‐0.012 0.007 1.198 0.257 0.031
0.007
1.417
0.299
‐0.009
0.008
1.575
0.299
0.030
0.008
0.401
0.268
‐0.056
0.005
1.656
0.124
‐0.012 0.003 0.781 0.156 0.075
0.028
2.140
0.792
272 0.251 0.528 233 0.143 0.438 272 0.305 0.568 272
0.175
0.439
272
0.233
0.467
272
0.017
0.313
272
0.484
0.630
272 0.163 1.002 243
0.090
0.246
Notes: OLS regression estimates; Newey-West standard errors. Top panel, 1986M08-1997M12, except for EU, 1999M01-2009M10.
Bottom panel, 1986M08-2009M10, except for NZ, 1989M01-2009M10. Entries in bold face denote significance at the 5% level, for
null hypothesis of beta=1.
18
Table 3.1 Bias, Three Month Horizon
coefficient constant gamma Obs. adj.R Sq. DW coefficient constant gamma Obs. adj.R Sq. DW BE FR GY IR
IT
NE
SP
EU ‐0.005 0.035 0.222 0.271 0.003 0.033 0.074 0.255 0.003 0.033 0.059 0.262 0.048
0.031
0.231
0.264
0.022
0.035
0.254
0.279
0.001
0.034
0.080
0.269
0.017
0.035
0.256
0.334
‐0.003 0.032 ‐0.091 0.131 141 0.003 0.548 140 ‐0.006 0.554 141 ‐0.006 0.558 141
0.006
0.572
141
0.002
0.482
141
‐0.006
0.562
141
0.004
0.662
128 ‐0.006 0.711 DE NO SN SW
UK
JP
AU
CA NZ
0.005 0.022 ‐0.038 0.145 0.019 0.030 ‐0.160 0.249 0.043 0.037 0.150 0.228 0.005
0.022
‐0.037
0.145
0.021
0.026
‐0.184
0.171
0.000
0.021
0.010
0.216
0.020
0.032
‐0.476
0.204
‐0.001 0.015 ‐0.154 0.251 0.034
0.023
0.107
0.074
272 ‐0.003 0.622 235 0.000 0.696 272 ‐0.002 0.456 272
‐0.003
0.600
272
0.001
0.610
272
‐0.004
0.560
272
0.017
0.933
272 0.000 0.848 272
0.034
0.584
Notes: OLS regression estimates; Newey-West standard errors. Top panel, 1986M08-1998M09, except for EU, 1999M01-2009M10.
Bottom panel, 1986M08-2009M10. Entries in bold face denote significance at the 5% level, for null hypothesis of gamma=1.
19
Table 3.2 Bias, Twelve Month Horizon
coefficient constant gamma Obs. adj.R Sq. DW coefficient constant gamma Obs. adj.R Sq. DW BE FR GY IR
IT
NE
SP
EU 0.010 0.021 ‐0.203 0.280 0.012 0.019 ‐0.255 0.258 0.002 0.019 ‐0.070 0.244 0.000
0.018
‐0.120
0.291
0.016
0.015
0.219
0.341
0.002
0.019
‐0.061
0.255
0.009
0.018
0.216
0.314
‐0.008 0.022 0.237 0.331 139 0.003 0.166 140 0.010 0.184 141 ‐0.006 0.173 141
‐0.004
0.165
139
0.005
0.156
141
‐0.006
0.166
139
0.001
0.134
128 0.005 0.163 DE NO SN SW
UK
JP
AU
CA NZ
‐0.006 0.013 0.052 0.182 ‐0.004 0.015 0.308 0.190 0.009 0.016 0.247 0.273 ‐0.012
0.013
0.055
0.167
‐0.003
0.012
0.311
0.228
‐0.017
0.013
‐0.036
0.236
‐0.007
0.016
‐0.055
0.266
‐0.006 0.011 0.425 0.393 0.017
0.015
0.551
0.148
272 ‐0.003 0.167 233 0.027 0.191 272 0.011 0.154 272
‐0.002
0.192
272
0.023
0.162
272
‐0.003
0.176
272
‐0.003
0.182
272 0.021 0.170 272
0.169
0.116
Notes: OLS regression estimates; Newey-West standard errors. Top panel, 1986M08-1997M12, except for EU, 1999M01-2009M10.
Bottom panel, 1986M08-2009M10. Entries in bold face denote significance at the 5% level, for null hypothesis of gamma=1.
20
Table 4. Panel Regressions
beta
Obs.
Adj.R sq.
Early
RatEx
3 Month
Early
Survey
3 Month
Full
RatEx
3 Month
0.913
(0.442)
1.255
(0.185)
1.106
(0.225
336
-0.006
335
0.179
828
0.021
Full
Survey
3 Month
Early
RatEx
12 Month
Early
Survey
12 Month
Full
RatEx
12 Month
Full
Survey
12 Month
0.318
(0.144
0.524
(0.408)
0.978
(0.215)
-0.220
(0.364)
1.221
(0.207)
815
0.062
84
-0.058
84
0.387
205
-0.036
204
0.146
Notes: OLS point estimates from fixed effects regressions. Non-overlapping data, regular standard errors. Early pertains to euro legacy
currencies, upper panels of Table 1 (except euro), from 1986m08-1998m09. Full pertains to currencies in lower panels of Table 1, from
1986m08-2009m10. Entries in bold face denote significance at the 5% level, for null hypothesis of beta=1.
21
Table 5. Persistence in the Risk Premium, Three Month Horizon
RatEx coefficient BE constant Survey BE RatEx FR Survey FR RatEx GY Survey GY RatEx IR Survey IR RatEx IT Survey IT RatEx NE Survey NE RatEx SP Survey SP RatEx EU Survey EU 0.002 ‐0.017 0.009 ‐0.018 0.009 ‐0.018 ‐0.049 0.004 0.003 ‐0.005 ‐0.001 ‐0.012 0.018 0.000 0.005 ‐0.006 0.028 0.014 0.028 0.014 0.036 0.017 0.029 0.012 0.030 0.010 0.029 0.011 0.031 0.012 0.028 0.018 0.213 0.546 0.182 0.543 0.082 0.310 0.170 0.576 0.143 0.627 0.142 0.654 0.176 0.648 0.150 0.390 0.102 0.106 0.109 0.095 0.104 0.244 0.099 0.099 0.110 0.075 0.090 0.073 0.114 0.088 0.110 0.041 48 48 48 48 48 48 48 48 48 48 48 48 48 48 43 43 adj.R Sq. 0.025 0.285 0.012 0.280 ‐0.015 0.077 0.008 0.314 ‐0.001 0.399 ‐0.001 0.422 0.010 0.408 0.001 0.129 Q(4)‐stat 6.804 2.728 6.875 3.194 3.258 5.055 6.431 0.944 5.464 4.327 8.378 2.744 1.188 5.464 2.594 0.801 p‐value 0.147 0.604 0.143 0.526 0.516 0.282 0.169 0.918 0.243 0.364 0.079 0.601 0.88 0.243 0.628 0.938 rho Obs. RatEx coefficient DE constant rho Obs. 0.011 Survey DE RatEx NO Survey NO 0.006 RatEx SN ‐0.010 Survey SN 0.002 RatEx SW ‐0.016 Survey SW ‐0.025 RatEx UK 0.002 Survey UK ‐0.001 RatEx JP Survey JP ‐0.031 ‐0.023 RatEx AU 0.012 Survey AU 0.018 RatEx CA 0.010 Survey CA RatEx NZ 0.012 ‐0.042 Survey NZ ‐0.007 0.015 ‐0.057 0.020 0.010 0.024 0.010 0.028 0.009 0.023 0.012 0.024 0.007 0.025 0.008 0.024 0.009 0.014 0.005 0.026 0.034 0.146 0.493 0.100 0.545 0.256 0.458 0.072 0.447 0.157 0.632 0.041 0.631 0.154 0.260 0.154 0.235 0.310 0.662 0.080 0.078 0.093 0.128 0.071 0.097 0.084 0.101 0.090 0.069 0.064 0.056 0.106 0.167 0.051 0.104 0.092 0.079 92 92 92 92 92 92 92 92 92 92 92 92 92 92 92 92 82 82 adj.R Sq. 0.011 0.228 ‐0.001 0.284 0.055 0.203 ‐0.006 0.191 0.014 0.392 ‐0.009 0.379 0.013 0.059 0.013 0.045 0.086 0.440 Q(4)‐stat 4.861 1.069 1.841 1.532 2.329 6.154 6.139 2.039 6.069 7.227 12.71 9.669 0.552 4.73 3.356 14.22 0.804 2.465 p‐value 0.302 0.899 0.765 0.821 0.675 0.188 0.189 0.729 0.194 0.124 0.013 0.046 0.968 0.316 0.5 0.007 0.938 0.651 Notes: Estimates from autoregression. Newey-West standard errors. Entries in bold face denote significance at the 5% level, for null hypothesis
of rho=0.
22
Figure 1
Three Month Risk Premia
.4
.4
.3
.3
.2
.2
.1
.1
.0
.0
-.1
-.1
-.2
-.2
-.3
-.3
-.4
-.4
-.5
-.5
86
87
88
89
90
91
92
RP3MBE
93
94
95
96
97
98
86
87
88
89
90
EPRP3MBE
91
92
RP3MFR
1.0
.4
0.8
.3
93
94
95
96
97
98
95
96
97
98
95
96
97
98
EPRP3MFR
.2
0.6
.1
0.4
.0
0.2
-.1
0.0
-.2
-0.2
-.3
-0.4
-.4
-0.6
-.5
86
87
88
89
90
91
92
RP3MGY
93
94
95
96
97
98
86
87
88
89
90
EPRP3MGY
91
92
RP3MIR
.4
93
94
EPRP3MIR
.4
.3
.2
.2
.0
.1
.0
-.2
-.1
-.4
-.2
-.3
-.6
-.4
-.8
-.5
86
87
88
89
90
91
RP3MIT
92
93
94
95
96
97
98
86
EPRP3MIT
87
88
89
90
91
RP3MNE
23
92
93
94
EPRP3MNE
.4
.4
.3
.2
.2
.1
.0
.0
-.2
-.1
-.2
-.4
-.3
-.6
-.4
-.5
-.8
86
87
88
89
90
91
92
93
RP3MSP
94
95
96
97
98
99
00
01
02
EPRP3MSP
03
04
RP3MEU
.4
0.8
.2
0.4
.0
0.0
-.2
-0.4
-.4
-0.8
-.6
-1.2
05
06
07
08
09
EPRP3MEU
-1.6
-.8
86
88
90
92
94
96
98
RP3MDE
00
02
04
06
86
08
88
90
92
94
96
98
RP3MNO
EPRP3MDE
0.4
00
02
04
06
08
04
06
08
EPRP3MNO
.6
.4
0.0
.2
-0.4
.0
-0.8
-.2
-1.2
-.4
-1.6
-.6
86
88
90
92
94
96
RP3MSN
98
00
02
04
06
08
86
EPRP3MSN
88
90
92
94
96
RP3MSW
24
98
00
02
EPRP3MSW
0.8
0.6
0.4
0.4
0.2
0.0
0.0
-0.4
-0.2
-0.4
-0.8
-0.6
-1.2
-0.8
-1.6
-1.0
86
88
90
92
94
96
98
RP3MUK
00
02
04
06
08
86
88
90
92
EPRP3MUK
94
96
98
RP3MJP
0.6
0.4
0.4
0.2
0.2
00
02
04
06
08
04
06
08
EPRP3MJP
0.0
0.0
-0.2
-0.2
-0.4
-0.4
-0.6
-0.6
-0.8
-0.8
-1.0
-1.0
86
88
90
92
94
96
RP3MAU
98
00
02
04
06
08
86
0.8
0.4
0.0
-0.4
-0.8
-1.2
-1.6
90
92
94
96
98
RP3MNZ
00
02
88
90
92
94
96
RP3MCA
EPRP3MAU
04
06
08
EPRP3MNZ
25
98
00
02
EPRP3MCA